The main motivation for population monitoring of large carnivores in Scandinavia is to inform adaptive management, i.e. to assess current state, evaluate the outcome of past decisions, and to guide future ones. Trial and error is not a viable option for large-scale management of wildlife, and there is a long history of using predictive population models to evaluate the outcome of different management scenarios (Peterson et al., 2003; Jepsen et al., 2005; McRae et al., 2008). Because the Bayesian OPSCR model developed during project RovQuant explicitly models population dynamics, it can be used not only to estimate past and current abundance, but also to forecast future population abundance under different management scenarios (e.g.
various harvest rates; Figure 24). In addition, the model captures the increase in uncertainty that arises as predictions are made further into the future (see violins in Figure 24). This application of the OPSCR model presents opportunities to evaluate the potential merits and risks of proposed management actions for the population. The next step in this development is an assessment of the reliability of forecasts made by the model by comparing predictions with estimates for years with known data. Conditional on the outcome of this assessment, the OPSCR models will be ready to use for making projections for alternative scenarios, at least for wolves and wolverines.
Season
Abundance
2017 2018 2019 2020 2021
0 200 400 600 800 1000
● ●
● ●
●
●
●
●
●
1
●
current culling mortality
●
2 x current culling mortality
●
0.5 x current culling mortality
Figure 24: Estimates and forecast of abundance for a subset of the Swedish wolverine population (male only).
Violins in the clear area show abundance estimates based on 3 years of non-invasive genetic sampling. Violins in the hashed area show abundance forecasts for two years for which non-invasive genetic sampling are not yet available. Forecasts are shown for three alternative levels of culling; “current” refers to the average annual hunting mortality estimated for the period with data (2017-2019).
4 General discussion
During project RovQuant we developed an OPSCR model that can be applied to monitoring data collected at an unprecedented scale. The model enabled us to exploit the extensive data collected by the Scandinavian monitoring programs and derive estimates of density, abundance and vital rates throughout the Scandinavian range of three large carnivore species: brown bears, wolverines, and wolves. We expect these results to become a valuable resource for large carnivore management, research, and public education in Scandinavia. Annual OPSCR-derived density maps for wolves have already been incorporated into the Scandinavian large carnivore database Rovbase 3.0 as an information source for managers and the public. Maps for bears and wolverines will soon be added.
RovQuant’s OPSCR model can now be applied to new monitoring data collected each year.
This will result in up-to-date abundance and density estimates, as well as a growing time series of these parameters. Because information is propagated through time in the model, annual estimates are expected to become more precise as the time series of available data lengthens.
Note that, as more information is incorporated, new estimates for earlier years may deviate from previously reported ones, especially for the bear, where the spatio-temporal configuration of sampling is not suited for range-wide estimation (subsection 3.4). In addition, continuation of the long-term monitoring and annual estimation will provide opportunities for testing and further development of the model. We strongly encourage an adaptive approach, whereby model predictions are validated against new information as it becomes available each year and necessary adjustments are made to the model to guarantee an evolution in both performance and reliability.
An essential step in the scientific process, is the rigorous vetting and review of ideas, methods, and results by the scientific community. Some of the work conducted as part of RovQuant has already been published in the peer-reviewed scientific literature, including versions of the OPSCR model (Bischof et al., 2016, 2017; Milleret et al., 2019b), specialized methods developed as part of the model (Milleret et al., 2018, 2019a), and tests of deviations from model assumptions (Dupont et al. et al., 2019, Bischof et al. in press). More of the work remains to be developed into scientific articles and subjected to peer review; this will be a focus for our research group during the coming two years.
With the ability to turn monitoring data compiled by the national monitoring programs in Sweden and Norway into annual estimates of spatially-explicit abundance using RovQuant’s OPSCR model, management is now nearing a crossroads with new options to monitor large carnivore populations, assess their status, and express population goals. In the current Norwe-gian management system, population goals are expressed as number of reproductions, i.e. the number of females (or, in the case of wolves, breeding packs) that have produced offspring. In Sweden, the population goals are expressed as number of individuals for all species, but even here, wolf abundance estimates are derived from the number of breeding packs using a conversion factor, Chapron et al. (2016); Bischof et al. (2019). The use of conversion factors or proxies for population size is in part a result of the elusiveness of direct estimates of abundance (Chapron et al., 2016). With both annual total and jurisdiction-specific population size estimates from the OPSCR model, management agencies in Scandinavia can now consider other options relying on direct carnivore abundance estimates in their management system, although this transition will be easier to implement in Sweden, where population goals are already expressed as abun-dance values. For brown bears in Sweden, we recommend a change to regular and range-wide sampling before relying on OPSCR-derived assessments of status in counties and years without sampling. Finally, in regions with very low abundance estimates (< 10 individuals), we advise against focusing on mean estimates, as the relative width of the credible interval is large and often includes or approaches zero.
RovQuant was made possible by the extensive long-term monitoring data compiled by the national monitoring programs in Sweden and Norway. At the same time, the project faced a
se-ries of significant challenges, due to the novelty of the methods employed and the unprecedented spatial scale at which RovQuant was working. Spatial capture-recapture models are a relatively recent development (Efford, 2004; Royle et al., 2014) and a series of conceptual and technical innovations were necessary to transition from the basic SCR model to an OPSCR model that included population dynamics, interannual movements, and multiple sources of information.
All developments were accompanied by extensive tests, including simulation studies, in order to ensure that the resulting models produced reliable estimates. Perhaps the most significant challenge was computational. SCR studies are typically conducted in study areas of small to moderate size, ranging from a few hectares to a few thousand square kilometers (e.g. Goldberg et al. 2015; Humm et al. 2017; Efford and Schofield 2019; Goswami et al. 2019; Nelson et al. 2019;
Petersen et al. 2019). By far, the largest study area used in SCR analysis outside of Scandinavia was a wolverine study in Canada (Mowat et al., 2019), with a total area of >50 000 km2and 126 detected individuals, analyzed with a single-season SCR model. The authors still had to take a number of shortcuts to reduce the size of the computational problem. By comparison, the sizes of the study areas involved in RovQuant were 526 000 km2 (bear), 593 000 km2 (wolverine), and 254 000 km2 (wolf), analyzed in a 7-year OPSCR model that included a total of 2 824 (bear), 2 118 (wolverine), and 1 092 (wolf) detected individuals. We were able to perform the millions of calculations required as part of these analyses (and the many tests and simulations conducted to test model performance) through methodological innovations that improved model efficiency (Milleret et al., 2018, 2019a), by implementing our modeling framework in NIMBLE with the assistance of the Nimble development team (NIMBLE Development Team, 2019; Turek et al., 2016; de Valpine et al., 2017), and by making use of computer clusters that allowed extensive parallelization. As a consequence, we are now able to fit OPSCR models to Scandinavian large carnivore monitoring data within a few days, whereas running earlier versions of our model would have not been feasible (e.g. using a rough extrapolation for a 2017-version of the model to this scale, we estimated a computation time of more than 3 years).
Aside from continuing the implementation and adaptation of the OPSCR model for ana-lyzing the annual large carnivore monitoring data, RovQuant’s work will be expanded on by WildMap, a research project funded by the Research Council of Norway (NFR 286886; 2019-2022; researchgate.net). WildMap focuses on further development of the computation methods for processing large scale monitoring data and on exploring the spatial patterns and drivers of wildlife population dynamics. Among other things, these developments should allow us to fur-ther improve the OPSCR models by incorporating more species-specific details in the population dynamics and observation process aspects of the model.
If population-level abundance estimation is a goal, we recommend that brown bear mon-itoring in Sweden transitions from the current configuration with shifting survey areas and multi-year gaps between consecutive sampling in any given region, to regular and range-wide monitoring. Although the cost-precision tradeoff of alternative sampling designs still needs to be assessed, it may be possible to maintain the current overall sampling intensity, but spread sampling to cover all or most of the brown bear range in Sweden and perhaps conduct sampling biennially or every third year. While range wide estimates would clearly be beneficial for a comprehensive assessment of population status, this tradeoff also includes the costs of reduced precision of estimates at the county level and the logistic challenges associated with mobiliz-ing volunteers for sample collection throughout a much larger spatial extent. Furthermore, we recommend avoiding non-random subsampling and prioritization of DNA samples subjected to genotyping, as this can invalidate the observation process model (function) and lead to biased estimates. This is especially a concern for wolves, where in most years resource and time con-straints lead to non-random selection of samples for analysis. Most importantly, we recommend continued open communications between management, monitoring, and research, to ensure cost efficient collection of data that is suitable for large scale analysis and yields reliable parameter estimates useful to managers and policy makers.
5 Acknowledgements
This work would not have been possible without the large carnivore monitoring programs and the extensive monitoring data collected by Swedish (Länstyrelsena) and Norwegian (SNO) wildlife management authorities, as well as the public in both countries. Our analyses relied on genetic analyses conducted by the laboratory personnel at the DNA laboratories at the Swedish University of Agricultural Sciences, Uppsala University, the Swedish Museum of Natu-ral History, the Norwegian Institute for Nature Research, and the Norwegian Institute of Bioe-conomy Research. We also thank Swedish and Norwegian wildlife managers for feedback they provided during the RovQuant workshop in June 2018 and the Swedish-Norwegian carnivore management meetings in December 2018 and 2019. The study was funded by Miljødirektoratet, Naturvårdsverket, and the Research Council of Norway (NFR 286886; project WildMap). The computations/simulations were performed on resources provided by NMBU’s computing cluster
“Orion”, administered by the Centre for Integrative Genetics and by UNINETT Sigma2 - the National Infrastructure for High Performance Computing and Data Storage in Norway. We are grateful to O. Gimenez and A. Royle for conceptual and methodological insights, and to P. de Valpine and D. Turek for extensive help with the formulation of the OPSCR model in Nimble.
We thank M. Tourani, M. Åkesson, L. Svensson, A. Ordiz, and G. Chapron for constructive discussions during the project period. J. Vermaat and V. Vasquez provided helpful comments on drafts of this report.
References
Åkesson, M., Liberg, O., Sand, H., Wabakken, P., Bensch, S., and Flagstad, Ø. (2016). Genetic rescue in a severely inbred wolf population.Molecular Ecology, 25(19):4745–4756.
Andreassen, R., Schregel, J., Kopatz, A., Tobiassen, C., Knappskog, P., Hagen, S., Kleven, O., Schneider, M., Kojola, I., Aspi, J., et al. (2012). A forensic dna profiling system for northern european brown bears (ursus arctos). Forensic Science International: Genetics, 6(6):798–809.
Armstrup, S., McDonald, T., and Manly, B. (2005). Handbook of Capture-Recapture Analysis. Princeton Univ.
Press, Princeton, New Jersey, USA.
Bischof, R. (2015). Approaches for assessing illegal hunting of brown bears and other large carnivores in sweden.
Report 2015:5 from the Scandinavian Brown Bear Research Project to the Swedish Environmental Protection Agency. Norwegian University of Life Sciences, Ås, Norway.
Bischof, R., Brøseth, H., and Gimenez, O. (2016). Wildlife in a politically divided world: Insularism inflates estimates of brown bear abundance. Conservation Letters, 9(2):122–130.
Bischof, R., Dupont, P., Milleret, C., Chipperfield, J., and Royle, J. A. ("in prep."). Consequences of ignoring group association in spatial capture-recapture analysis. Wildlife Biology.
Bischof, R., Milleret, C., Dupont, P., Chipperfield, J., Åkesson, M., Brøseth, H., and Kindberg, J. (2019).
Estimating the size of the scandinavian wolf population with spatial capture recapture and conversion factors.
MINAfagrapport57, 80pp.
Bischof, R., Steyaert, S. M. J. G., and Kindberg, J. (2017). Caught in the mesh: roads and their network-scale impediment to animal movement.Ecography, 40(12):1369–1380.
Bischof, R., Swenson, J. E., Yoccoz, N. G., Mysterud, A., and Gimenez, O. (2009). The magnitude and selectivity of natural and multiple anthropogenic mortality causes in hunted brown bears. Journal of Animal Ecology, 78(3):656–665.
Borchers, D. L. and Efford, M. G. (2008). Spatially explicit maximum likelihood methods for capture-recapture studies. Biometrics, 64(2):377–385.
Brøseth, H., Flagstad, ., Wärdig, C., Johansson, M., and Ellegren, H. (2010). Large-scale noninvasive genetic mon-itoring of wolverines using scats reveals density dependent adult survival. Biological Conservation, 143(1):113 – 120.
Chapron, G., Wikenros, C., Liberg, O., Wabakken, P., Flagstad, Ø., Milleret, C., Månsson, J., Svensson, L., Zimmermann, B., Åkesson, M., and Sand, H. (2016). Estimating wolf (canis lupus) population size from number of packs and an individual based model. Ecological Modelling, 339:33 – 44.
Dahle, B. and Swenson, J. E. (2003). Home ranges in adult scandinavian brown bears (ursus arctos): effect of mass, sex, reproductive category, population density and habitat type. Journal of Zoology, 260(4):329–335.
Dawson, F. N., Magoun, A. J., Bowman, J., and Ray, J. C. (2010). Wolverine, gulo gulo, home range size and denning habitat in lowland boreal forest in ontario. The Canadian Field-Naturalist, 124(2):139–144.
de Valpine, P., Turek, D., Paciorek, C. J., Anderson-Bergman, C., Lang, D. T., and Bodik, R. (2017). Pro-gramming with models: writing statistical algorithms for general model structures with nimble. Journal of Computational and Graphical Statistics, 26(2):403–413.
Dupont, P., Milleret, C., Gimenez, O., and Bischof, R. (2019). Population closure and the bias-precision trade-off in spatial capture–recapture. Methods in Ecology and Evolution, 10(5):661–672.
Efford, M. (2004). Density estimation in live-trapping studies.Oikos, 106(3):598–610.
Efford, M. G. (2011). Estimation of population density by spatially explicit capture-recapture analysis of data from area searches. Ecology, 92(12):2202–2207.
Efford, M. G. and Schofield, M. R. (2019). A spatial open-population capture-recapture model. Biometrics.
Ergon, T. and Gardner, B. (2014). Separating mortality and emigration: modelling space use, dispersal and survival with robust-design spatial capture-recapture data. Methods in Ecology and Evolution, 5(12):1327–
1336.
Flagstad, Ø., Hedmark, E., Landa, A., Brøseth, H., Persson, J., Andersen, R., Segerstrøm, P., and Ellegren, H.
(2004). Colonization history and noninvasive monitoring of a reestablished wolverine population.Conservation Biology, 18(3):676–688.
Flagstad, Ø., Kleven, O., Spets, M., Eriksen, L., Halvorsen, B., Andersskog, I., ..., and Brøseth, H. (2018). Dna-based monitoring of the scandinavian wolverine population 2018. NINA Report 1592. Norwegian Institute for Nature Research.
Fløystad, I., Brøseth, H., Bakke, B., Eiken, H., and Hagen, S. (2019). Populasjonsovervåking av brunbjørn.
dna-analyse av prøver innsamlet i norge i 2018. NINA Rapport 1658.Norsk institutt for naturforskning.
Gardner, B., Sollmann, R., Kumar, N. S., Jathanna, D., and Karanth, K. U. (2018). State space and movement specification in open population spatial capture-recapture models. Ecology and Evolution, 8(20).
Gelman, A. and Rubin, D. (1992). Inference from iterative simulation using multiple sequences.Statistical Science, 7:457–511. http://www.stat.columbia.edu/~gelman/research/published/itsim.pdf.
Goldberg, J. F., Tempa, T., Norbu, N., Hebblewhite, M., Mills, L. S., Wangchuk, T. R., and Lukacs, P. (2015).
Examining temporal sample scale and model choice with spatial capture-recapture models in the common leopard panthera pardus. PLOS ONE, 10(11):1–19.
Goswami, V. R., Yadava, M. K., Vasudev, D., Prasad, P. K., Sharma, P., and Jathanna, D. (2019). Towards a reliable assessment of asian elephant population parameters: the application of photographic spatial capture–
recapture sampling in a priority floodplain ecosystem. Scientific reports, 9(1):8578.
Gyllenstrand, N. (2018). Björnspillningsinventering 2017. rapport från naturhistoriska riksmuseet. Centrum för genetisk identifiering. 2018-06-26, Diarienr 4.1-17-2017.
Höglund, L., Mattisson, J., and Tovmo, M. (2019). Bestandsovervåking av jerv i 2019. bestandsstatus for store rovdyr i skandinavia. Beståndsstatus för stora rovdjur i Skandinavien.
Humm, J. M., McCown, J. W., Scheick, B. K., and Clark, J. D. (2017). Spatially explicit population estimates for black bears based on cluster sampling. The Journal of Wildlife Management, 81(7):1187–1201.
Jepsen, J., Topping, C., Odderskaer, P., and Andersen, P. (2005). Evaluating consequences of land-use strategies on wildlife populations using multiple-species predictive scenarios. Agriculture, ecosystems & environment, 105(4):581–594.
Kindberg, J. and Swenson, J. E. (2014). Björnstammens storlek i sverige 2013 länsvisa skattningar och trender.
Scandinavian Brown Bear Research Project report.
Kindberg, J. and Swenson, J. E. (2018). Björnstammens storlek i sverige 2017.Scandinavian Brown Bear Research Project report.
Kindberg, J., Swenson, J. E., Ericsson, G., Bellemain, E., Miquel, C., and Taberlet, P. (2011). Estimating population size and trends of the swedish brown bear ursus arctos population.Wildlife Biology, 17(2):114–123.
Liberg, O., Aronson, Å., Sand, H., Wabakken, P., Maartmann, E., Svensson, L., and Åkesson, M. (2012). Moni-toring of wolves in scandinavia. Hystrix, the Italian Journal of Mammalogy, 23(1):29–34.
Liberg, O., Chapron, G., Wabakken, P., Pedersen, H. C., Hobbs, N. T., and Sand, H. (2011). Shoot, shovel and shut up: cryptic poaching slows restoration of a large carnivore in europe. Proceedings of the Royal Society B:
Biological Sciences, 279(1730):910–915.
Luikart, G., Ryman, N., Tallmon, D. A., Schwartz, M. K., and Allendorf, F. W. (2010). Estimation of census and effective population sizes: the increasing usefulness of dna-based approaches. Conservation Genetics, 11(2):355–373.
Mattisson, J., Persson, J., Andren, H., and Segerström, P. (2011). Temporal and spatial interactions between an obligate predator, the eurasian lynx (lynx lynx), and a facultative scavenger, the wolverine (gulo gulo).
Canadian Journal of Zoology, 89(2):79–89.
Mattisson, J., Sand, H., Wabakken, P., Gervasi, V., Liberg, O., Linnell, J., Rauset, G., and Pedersen, H. (2013).
Home range size variation in a recovering wolf population: evaluating the effect of environmental, demographic, and social factors. Oecologia, 173(3):1–13.
Mattisson, J., Støen, O.-G., and Kindberg, J. (2019). Seasonal movement of brown bears in relation to sex and age groups. NINA Report 1662. NINA Report 1662. Norwegian Institute for Nature Research.
McRae, B. H., Schumaker, N. H., McKane, R. B., Busing, R. T., Solomon, A. M., and Burdick, C. A. (2008). A multi-model framework for simulating wildlife population response to land-use and climate change. ecological modelling, 219(1-2):77–91.
Milleret, C., Dupont, P., Bonenfant, C., Brøseth, H., Flagstad, y., Sutherland, C., and Bischof, R. (2019a). A local evaluation of the individual state-space to scale up bayesian spatial capture recapture.Ecology and Evolution, 9(1):352–363.
Milleret, C., Dupont, P., Brøseth, H., Kindberg, J., Royle, J. A., and Bischof, R. (2018). Using partial aggregation in spatial capture recapture. Methods in Ecology and Evolution, 9(8):1896–1907.
Milleret, C., Dupont, P., Chipperfield, J., Turek, D., Brøseth, H., Gimenez, O., de Valpine, P., and Bischof, R.
(2019b). Estimating abundance with interruptions in data collection using open population spatial capture-recapture models. bioRxiv.
Mowat, G., Clevenger, A. P., Kortello, A. D., Hausleitner, D., Barrueto, M., Smit, L., Lamb, C., DorsEy, B., and Ott, P. K. (2019). The sustainability of wolverine trapping mortality in southern canada. The Journal of Wildlife Management.
Nelson, D. L., Kellner, K. F., and Swihart, R. K. (2019). Rodent population density and survival respond to disturbance induced by timber harvest. Journal of Mammalogy, 100(4):1253–1262.
NIMBLE Development Team (2019).NIMBLE: MCMC, Particle Filtering, and Programmable Hierarchical Mod-eling. https://cran.r-project.org/package=nimble.
Persson, J., Ericsson, G., and Segerström, P. (2009). Human caused mortality in the endangered scandinavian wolverine population. Biological conservation, 142(2):325–331.
Petersen, W. J., Savini, T., Steinmetz, R., and Ngoprasert, D. (2019). Estimating leopard cat prionailurus bengalensis kerr, 1792 (carnivora: Felidae) density in a degraded tropical forest fragment in northeastern thailand. Journal of Threatened Taxa, 11(4):13448–13458.
Peterson, G. D., Cumming, G. S., and Carpenter, S. R. (2003). Scenario planning: a tool for conservation in an uncertain world. Conservation biology, 17(2):358–366.
Plummer, M. (2003). Jags: A program for analysis of bayesian graphical models using gibbs sampling. In Proceedings of the 3rd international workshop on distributed statistical computing, volume 124, page 125. Vienna.
R Core Team (2018). R: A language and environment for statistical computing. vienna, .Austria: R Foundation for Statistical Computing.
Royle, J. A., Chandler, R. B., Sollmann, R., and Gardner, B. (2014).Spatial Capture-Recapture. Academic Press.
Royle, J. A., Karanth, K. U., Gopalaswamy, A. M., and Kumar, N. S. (2009). Bayesian inference in camera trapping studies for a class of spatial capture-recapture models. Ecology, 90(11):3233–3244.
Royle, J. A. and Young, K. V. (2008). A hierarchical model for spatial capture–recapture data. Ecology, 89(8):2281–2289.
Squires, J. R., Copeland, J. P., Ulizio, T. J., Schwartz, M. K., and Ruggiero, L. F. (2007). Sources and patterns of wolverine mortality in western montana.The Journal of Wildlife Management, 71(7):2213–2220.
Svensson, L., Wabakken, P., Maartmann, E., Åkesson, M., Flagstad, ., and Hedmark, E. (2019). Inventering av varg vintern 2018-2019. bestandsovervåking av ulv vinteren 2018-2019. Bestandsstatus for store rovdyr i Skandinavia. Beståndsstatus för stora rovdjur i Skandinavien, 1(53s).
Swenson, J. E., Sandegren, F., Brunberg, S., Segerström, P., and Segerstrøm, P. (2001). Factors associated with loss of brown bear cubs in sweden. Ursus, pages 69–80.
Taberlet, P., Griffin, S., Goossens, B., Questiau, S., Manceau, V., Escaravage, N., Waits, L. P., and Bouvet, J. (1996). Reliable genotyping of samples with very low dna quantities using pcr. Nucleic acids research, 24(16):3189–3194.
Tobiassen, C., Brøseth, H., Bergsvåg, M., Aarnes, S. G., Bakke, B. B., Hagen, S., and Eiken, H. G. (2011).
Tobiassen, C., Brøseth, H., Bergsvåg, M., Aarnes, S. G., Bakke, B. B., Hagen, S., and Eiken, H. G. (2011).